Perplexity AI

How It Works

Perplexity runs a live web search for your query first, then feeds the retrieved pages to an LLM as source material - a pattern called retrieval-augmented generation. The model writes an answer grounded in those specific pages and attaches numbered citations linking each claim back to its source, rather than answering purely from what it learned during training. Deep Research repeats this search-read-synthesize loop many times over to produce a longer, multi-source report.

Key Features